Speech to Text

Version 0.9.12 built on 2026-09-02

Download for Windows x64 16.8 MB
Portable Windows version 13.5 MB
Installation guide

This program requires model files to work; the recommended model is 3.4 GB. A downloader is included.
The installer usually asks for administrator rights. Approving the prompt is optional – see the installation guide for details.

Version 0.9.12 RC5

Third GPU backend based on Vulkan 1.3 in addition to D3D11 and D3D12 which were already there.

Shader binary cache for D3D12 and Vulkan backends, stored in %LOCALAPPDATA%/Speech to Text/PipelineCache.zip

Advanced option (disabled by default) to produce diagnostic log text files.

Advanced option (disabled by default) to write GPU crash dumps when shaders fail on nVidia GPUs.

On fatal exceptions, applicable apps now save a minidump (enabled by default), and a companion tool can upload crash reports to the server.
That tool never uploads anything automatically; it requires a click on the upload button. The uploaded reports are encrypted locally using strong asymmetric cryptography: the server does not have the decryption key.

Benchmark tool: added an option (enabled by default) to detect when GPU throughput drifts due to power state or thermal budget shenanigans. When that happens, the tool reruns the completed tests.

A few UX and performance improvements.

SHA-512 of the setup.exe: a0980aa4 … 0db210b6
a0980aa43c95f5c54ccefa6bf6e7bdf9a7bc08378417e9ec7ceacb20141926edef9eac0027dd6a1ec3f4649cfa23ab02b2d8b17a5c3ace95e0db65d90db210b6